85 min. What is a JITAI -- how it relates to multi-component interventions and to AIs Motivation for JITAI What is MRT (basic features) The connection between MRT and MOST, Factorial designs and SMART

SA Murphy 1 three available stress-regulation apps (headspace; mood-surfing; thought distancing; and cognitive restructuring).

2 Sedentary adults Next study will be with adults who are at high risk for heart attack

3 4 5 Monitoring, individualizing, delivering

6 The same elements that we used to describe an adaptive intervention can be used to describe JITAISs, only that now all these elements are in-the-.

Dynamic Models of Behavior for Just-in-Time Adaptive Interventions Donna Spruijt-Metz, University of Southern California Wendy Nilsen, National Institutes of Health PERVASIVE computing, 1536-1268/14/2014 IEEE

Nahum-Shani, I., Smith, S.N. Spring, B.J., Collins, L.M., Witkiewitz, K., Tewari, A., & Murphy, S.A. . (2016). Just-in-Time Adaptive Interventions (JITAIs) in Mobile Health: Key Components and Design Principles for Ongoing Health Behavior Support. Annals of Behavioral Medicine . doi:10.1007/s12160-016-9830-8, PMCID: PMC5364076

Different terms have been used in various fields to describe a JITAI, including dynamic tailoring, intelligent real-time therapy, and dynamically and individually tailored EMI

7 Development and evaluation of a mobile intervention for heavy drinking and smoking among college students. Witkiewitz, Katie; Desai, Sruti A.; Bowen, Sarah; Leigh, Barbara C.; Kirouac, Megan; Larimer, Mary E. Psychology of Addictive Behaviors, Vol 28(3), Sep 2014, 639-650

8 Quote from paper: whenever 30 min of nearly uninterrupted computer activity was recorded, a short text message (SMS) containing a hyperlink was sent to the participant’s smart phone. When participants clicked on this hyperlink, they were shown a message persuading them to be more active. Although all messages contained the same general advice, this advice was phrased in various ways, using four different persuasive strategies. The four strategies are a subset of the six social influence strategies defined by Cialdini [22]. Dantzig, S., Geleijnse, G., & Halteren, A. T. (2013). Toward a persuasive mobile application to reduce sedentary behavior. Personal and ubiquitous computing, 17 (6), 1237-1246.

9 The same elements that compose an adaptive intervention, also compose a JITAI. However, in a JITAI these elements are in-the-moment – they can occur at any moment.

10 11 12 In MD2K smoking study the distal outcome might be time to relapse.

13 Likely multiple proximal outcomes

In MD2k study the proximal outcome might be stress over next x minutes.

14 Intervention options are typically designed to impact distal outcome via the proximal outcome.

In MD2K study the intervention option might be a recommendation to access one of the three stress-regulation apps (headspace; mood-surfing; and ?) residing on the smartphone vs. no recommendation.

Intervention options in JITAIs include types of support, sources of support (e.g., automated sources, social sources); and modes of support delivery. Recommendations Reach out recommendation (contact a friend)

15 Behavioral strategies (exercise; stay in locations that are supportive of change) Cognitive strategies (relaxation; reframing) Motivational messages (reasons for behavior change; barriers for change); Setting goals; modifying goals Feedback (often with visualization: fish; flower; garden) Distractions (game, music, etc.)

Michel Klein et al. have a nice review of health behavior change theories used to inform EMIs Klein, M., Mogles, N., & van Wissen, A. (2011). Why won’t you do what’s good for you? Using intelligent support for behavior change. In Human Behavior Understanding (pp. 104-115). Springer Berlin Heidelberg.

West, R., & Michie, S. (2016). A Guide to Development and Evaluation of Digital Interventions in Healthcare. London: Silverback In MD2K smoking study tailoring variables might be current classification of stressed or not and location (home, work), time of day (before work, during work, after work). Also weather. indicate risk or vulnerability. --internal risk factors , external risk factors: behaviors, social context, geographical location, When user ignores assessment requests or ignores intervention

16 Recall that a decision point is the time in which we need to make critical decisions about the intervention options based on patient information. decision points can result in the “do nothing intervention option,” hence a decision point every 3 minutes does not imply an intervention every 3 minutes.

17 The decision rules are constructed with the aim to impact a proximal outcome.

We can use the data from the micro-randomized study along with behavioral science to construct decision rules.

18 19 20 21 Make a break –everyone stretches!

22 Sedentary adults Next study will be with adults who are at high risk of an adverse cardiac event

23 three available stress-regulation apps (headspace; mood-surfing; thought distancing; and cognitive restructuring).

Currently, AutoSense consists of an arm band with four wireless sensors and a chestband with six wireless sensors. All the ten sensors are integrated onto an embedded platform called “mote,” a tiny self-contained, battery-powered computer with a wireless radio that can host multiple sensors, collect and process data from them using customized algorithms, and communicate on secure wireless channels.

The chestband consists of 2-lead Electrocardiogram (ECG), galvanic skin response (GSR), respiratory inductive plethysmograph (RIP) band for robust measurement of respiration, skin temperature, ambient temperature, and a 3-axis accelerometer. Accelerometers help classify physical activities, estimate their intensities, and help remove motion artifacts from the measurements of ECG, RIP, and GSR. The armband consists of WrisTAS alcohol sensor from Giner Inc., accelerometers, GSR, and temperature sensors. All sensors communicate wirelessly with a smart phone. Sensors on the phone (e.g., GPS, microphone) complement those on the body. The phone also acts as a local server for heavier computation and storage.

A person is available if he/she (1) is wearing autosense (our understanding is that autosense will be worn up to 16 hours a day, participants will not wear it when they

24 sleep or when in the shower; if the person is not wearing Autosense, no data will be collected and recommendations will not be pushed); (2) did not receive a message in the past x minutes; and (3) is not driving a car. SA Murphy 25 The momentary times were selected because these times are the times at which most people are able to be active Pre-morning commute, mid-day, mid-afternoon, evening commute, after dinner.

Another example: The phone software monitors a risk measure at regular time intervals and if the risk measures hits a criterion then a treatment is provided. Can include time of day or day of week and present weather. Some Treatment types behavioral, cognitive, motivational, social, self-monitoring, information Tailored to time of day, location, weather outside, weekday

The location of the like button biases against the person hitting like. The snozz button turns off the momentary lock screen recommendations for x hours.

Occurs up to 5 times per day

SA Murphy 29 Frequently the actions are primarily designed to have a near-term effect on the individual. E.g. Help then manage current craving/stress, help them manage or be aware of the impact of their social setting on their craving/stress 31 32 33 34 J=42*5=210 momentary randomizations

35 36 Availability is measured prior to randomization. Adherence is measured after randomization.

Adherence (i.e. compliance) is very different from availability. Suppose a person is available at a decision point. However the phone is in their purse across the room. So they don’t hear whether the phone pings/ see the lockscreen light up. This person is non-adherent at this decision point. Primary analyses will be intention-to-treat and thus will average over non-adherence.

SA Murphy 37 Marginal over randomization (and effects thereof), conditional on those who are available.

The group who are available is a selected group of people likely depending on the intervention dose they experienced up to time j. This intervention dose may have caused burden, may have caused learning.

SA Murphy 38 In heartsteps beta(t) is the effect of tailored activity suggestion on next 30min step count.

Delayed effects which are akin to higher order interactions would be investigated in secondary analyses

SA Murphy 39 Instead of a sparsity bet, we place a smoothness bet We are not assuming that the main effect has a quadratic form

Since the test for the main effect does not depend on time of day, we are averaging any variation in main effect across the occasions during the day (recall we are sizing the study; a primary analysis might be a little more complex and in secondary data analyses one would likely estimate and test if the effect varies by time of day and/or varies by day in study).

SA Murphy 40 SA Murphy 41 The contrasts become within person contrasts due to smoothness over time in the targeted quadratic alternative. If the main effect at each time point were to be estimated separately then it would be like a two arm study at each time j.

SA Murphy 42 SA Murphy 43 Meaningful increase in stepcount is 1000/day Usual std is 2000/day Roughly a standardized treatment effect of 200/666= .3

Sample size calculators https://pengliao.shinyapps.io/mrt-calculator/ R package https://cran.r-project.org/web/packages/MRTSampleSize/index.html

SA Murphy 44 45 46 47 48 Single-patient trials or individual-patient trials or single-case

49 50 ABBA trials, single case designs, quasi-experimental

Design and Implementation of N-of-1 Trials: A users’ guide https://www.effectivehealthcare.ahrq.gov/ehc/products/534/1844/n-1-trials-report- 130213.pdf

Kravitz RL, Duan N, eds, and the DEcIDE Methods Center N-of-1 Guidance Panel (Duan N, Eslick I, Gabler NB, Kaplan HC, Kravitz RL, Larson EB, Pace WD, Schmid CH, Sim I, Vohra S). Design and Implementation of N-of-1 Trials: A User’s Guide. AHRQ Publication No. 13(14)-EHC122-EF. Rockville, MD: Agency for Healthcare Research and Quality; January 2014. http://www.effectivehealthcare.ahrq.gov/N-1-Trials.cfm.

51 No notion of availability 53 This project tests the feasibility and effectiveness of providing, via a smartphone, just-in-time tailored physical activity suggestions as well as evening prompts to plan the following day's physical activity so as to help sedentary individuals increase their activity. The resulting data will be used to inform the development of a JITAI for increasing physical activity. PI: Predrag Klasnja Location: Funding: NHLBI/NIA R01HL125440 heartsteps MRT https://www.clinicaltrials.gov/ct2/show/NCT03225521?titles=HeartSteps&rank=1

54 55 This project tests the feasibility of conducting an MRT aiming to investigate whether real-time sensor-based assessments of stress are useful in optimizing the provision of just-in-time prompts to support stress-management in chronic smokers attempting to quit. The resulting data will be used to inform the development of a JITAI for smoking cessation. PI: Santosh Kumar, Center of Excellence for Mobile Sensor Data-to-Knowledge (MD2K, https://md2k.org ) Location: Northwestern University, Bonnie Spring, (site P.I.) Funding: NIBIB through funds provided by the trans-NIH Big Data to Knowledge (BD2K) initiative (www.bd2k.nih.gov ). U54EB020404

MD2K smoking cessation study https://www.clinicaltrials.gov/ct2/show/study/NCT03184389?recrs=a&lead=North western+University&cntry1=NA%3AUS&state1=NA%3AUS%3AIL&draw=1

56 SA Murphy 57 58 Researchers are conducting this quality-improvement MRT aiming to promote weight maintenance through increased activity and improved diet among people who received bariatric surgery. At the time it was developed, this project was novel in that it implemented separate randomizations at the start of the study, on a daily basis, and five times throughout the day. PI: Pedja Klasna Location & Funding: Kaiser Permanente 50 participants in a 4-month Micro-randomized trial. Before the study: There is a 7-day baseline prior to randomization. Participants will be wearing a screen-less accelerometer for a week before they start the study, to capture step that will be used to generate adaptive step goals ( values).

59 Rabbi, M., Philyaw-Kotov, M., Klasnja, P., Bonar, E., Nahum-Shani, I., Walton, M., & Murphy, S. (2017, October 18). SARA - Substance Abuse Research Assistant. Retrieved from osf.io/vwzmd Go to the files section.

SA Murphy 60

The Substance Abuse Research Assistance (SARA) is an app for gathering data about substance use in high-risk populations. App developers are using an MRT to improve engagement with completion of the self-report data collection measures. At the time this summary was written, this MRT is unique in that it has an engagement component, but not a treatment one. PIs: Maureen Walton, Susan Murphy, and Mashfiqui Rabbi Shuvo Location: and University of Michigan Funding: Michigan Institute for Data Science (PI S. Murphy), the University of Michigan Injury Center (PI M. Walton), NIDA P50 DA039838 (PI Linda Collins), NIAAA R01 AA023187 (PI S. Murphy), CDC R49 CE002099 (PI: M. Walton) https://clinicaltrials.gov/ct2/show/NCT03255317 And https://osf.io/whgfp/

62 JOOL is a behavioral health and well ‐being app that is designed to help people monitor and improve their sleep, presence, activity, creativity, and eating, with the ultimate goal of helping people move closer to fulfilling their life’s purpose. This MRT aims to understand whether push notifications of tailored health messages are useful in promoting engagement with the JOOL app; and, if so, when and under what circumstances they are most effective. PI: Vic Strecher, PhD, MPH, CEO of JOOL Health Location: Ann Arbor, Michigan URL: https://www.joolhealth.com

63